1. Field of the Invention
The present invention relates to an image processing apparatus and an image processing method for subjecting an input color image to image processing, for example, in a digital color copying machine that reads a color image on an original and forms a copy image of the color image.
2. Description of the Related Art
In an apparatus, such as a digital copying machine, which handles a document image as a digital signal, in order to perform image recording with both high resolution and a good tone rendition, an image region discrimination process is used.
In the image region discrimination process, an image to be recorded is discriminated into a part, e.g. a photo part, for which a tone rendition is important, and a part, e.g. a text part, for which resolution is important. Based on the discrimination result, a recording process is switched.
Image region discrimination methods, in general terms, fall into three categories: microscoped-discrimination which uses a difference in local characteristic such as microscopic image density or density variation; macro-scoped correction which makes fine corrections of a discrimination result by referring to a micro-discrimination result in a large region of a certain size; and macro-discrimination which discriminates a region on the basis of a result of general analysis of document structure.
For example, in conventional micro-discrimination techniques disclosed in Jpn. Pat. Appln. KOKAI Publication No. 58-3374 and Jpn. Pat. Appln. KOKAI Publication No. 60-204177, the presence/absence of an edge part is basically discriminated. It is thus difficult to increase the precision in discrimination between an edge part of a tone image and a character part, or in discrimination of an inside of a thick character.
In the macro-scoped discrimination, micro-discrimination is first performed and then correction is made by referring to a discrimination result of a region near a pixel of interest, making use of the feature that an image region, such as a halftone-screen region or a photo region, is constant in a relatively broad range. With this method, the precision in discrimination can be improved. However, if a large region is set as a reference region, the circuitry scale increases, leading to a rise in cost. In addition, since an edge part is basically detected, it is difficult to enhance the precision in discrimination between an edge part of a tone image and an inside part of a thick character.
For example, in the conventional macro-discrimination technique disclosed in, e.g. “Extraction algorithm of character region for mixed-mode communication”, IEICE (the Institute of Electronics, Information and Communication Engineers), Trans., J67-D, vol. 11, pp. 1277-1284 (1984), a document structure is analyzed from information on an entire document image. Thus, an enormous length of processing time is required, or a high-speed processor is required.